CodeAttack/substitues.py at main · reddy-lab-code-research/CodeAttack · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
import os, sys
parent_dir = os.path.abspath(os.path.dirname(__file__))
sys.path.append(parent_dir)

import torch
import torch.nn as nn
import torch.nn.functional as F

import pandas as pd


def constrain_type(config, tgt_word, final_words):
    in_file = os.path.join(config['output_dir'], config['victim_model'] + '.txt')
    fp = open(in_file, 'w+')
    fp.write(tgt_word.strip() + '\n')
    for words in final_words:
        # Tokenization cannot handle nested "'"
        if "'" in words:
            words = words.replace("'", "")
        fp.write(words.strip() + '\n')
    fp.close()

    out_file = os.path.join(config['output_dir'], config['victim_model'] + '.json')
    script_file = '/home/akshitajha/adversarial_attack/scripts/tokenizer/./tokenize'
    command =  script_file + ' ' + in_file + ' -m csv -o ' + out_file
    os.system(command)

    # Saved in the format [line, column, class, token]
    data = pd.read_csv(out_file)

    meta_data = {}
    token_class = ['identifier', 'keyword', 'integer', 'floating', 'string', 'character', 'operator', 'preprocessor', 'sum_classes']

    # Insert target word as the first word
    # final_words indexes from 0
    # meta_data indexes from 1
    final_words.insert(0, tgt_word)
    for i in range(1, len(final_words)+1):
        meta_data[i] = {}
        for c in token_class:
            meta_data[i][c] = 0

    for idx, d in data.iterrows():
        meta_data[d['line']][d['class']] =  meta_data[d['line']][d['class']] + 1
        meta_data[d['line']]['sum_classes'] = meta_data[d['line']]['sum_classes'] + 1

    tgt_class = meta_data[1]['sum_classes']
    poss_words = []
    rejected_words = []

    # Stricter Type Constraint
    tgt_tokens = data[data['line']==1]['class'].values.tolist()

    # If DFG constraint only
    if config['use_dfg_constraint'] == 1:
        if list(set(tgt_tokens)) != ['operator']:
            return poss_words

    for key in meta_data:
        # Making sure the same classes are replaced
        if meta_data[key]['sum_classes'] == tgt_class:
            sub_tokens = data[data['line']==key]['class'].values.tolist()
            if tgt_tokens == sub_tokens:
                poss_words.append(final_words[key-1])
            else:
                rejected_words.append(final_words[key-1])

        # Addition of an operator
        elif meta_data[key]['sum_classes'] == tgt_class + 1:
            sub_tokens = data[data['line']==key]['class'].values.tolist()
            sub_len = sub_tokens.count('operator')
            tgt_len = tgt_tokens.count('operator')
            if sub_len - tgt_len == 1:
                # Make sure the token classes are the same except an operator
                # A keyword is only replaced by a keyword, etc.
                if set(sub_tokens) == set(tgt_tokens):
                    poss_words.append(final_words[key-1])
                else:
                    rejected_words.append(final_words[key-1])
            else:
                rejected_words.append(final_words[key-1])

        # Deletion of an operator
        elif meta_data[key]['sum_classes'] == tgt_class - 1:
            sub_tokens = data[data['line']==key]['class'].values.tolist()
            sub_len = sub_tokens.count('operator')
            tgt_len = tgt_tokens.count('operator')
            if tgt_len - sub_len == 1:
                # Make sure the token classes are the same except an operator
                # A keyword is only replaced by a keyword, etc.
                if set(sub_tokens) == set(tgt_tokens):
                    poss_words.append(final_words[key-1])
                else:
                    rejected_words.append(final_words[key-1])
            else:
                rejected_words.append(final_words[key-1])
        else:
            rejected_words.append(final_words[key-1])

    return poss_words[1:]


def get_bpe_substitues(config, tgt_word, substitutes, tokenizer, mlm_model):
    # substitutes L, k

    substitutes = substitutes[0:12, 0:4] # maximum BPE candidates

    # find all possible candidates

    all_substitutes = []
    for i in range(substitutes.size(0)):
        if len(all_substitutes) == 0:
            lev_i = substitutes[i]
            all_substitutes = [[int(c)] for c in lev_i]
        else:
            lev_i = []
            for all_sub in all_substitutes:
                for j in substitutes[i]:
                    lev_i.append(all_sub + [int(j)])
            all_substitutes = lev_i

    # all substitutes  list of list of token-id (all candidates)
    c_loss = nn.CrossEntropyLoss(reduction='none')
    word_list = []
    # all_substitutes = all_substitutes[:24]
    all_substitutes = torch.tensor(all_substitutes) # [ N, L ]
    all_substitutes = all_substitutes[:24].to('cuda')
    # print(substitutes.size(), all_substitutes.size())
    N, L = all_substitutes.size()
    word_predictions = mlm_model(all_substitutes)[0] # N L vocab-size
    ppl = c_loss(word_predictions.view(N*L, -1), all_substitutes.view(-1)) # [ N*L ]
    ppl = torch.exp(torch.mean(ppl.view(N, L), dim=-1)) # N
    _, word_list = torch.sort(ppl)
    word_list = [all_substitutes[i] for i in word_list]
    final_words = []
    for word in word_list:
        tokens = [tokenizer._convert_id_to_token(int(i)) for i in word]
        text = tokenizer.convert_tokens_to_string(tokens)
        final_words.append(text)

    # Add Type Detection as a constraint
    if config['use_ast_constraint'] == 1:
        final_words = constrain_type(config, tgt_word, final_words)

    return final_words


# def get_substitues(config, tgt_word, substitutes, tokenizer, mlm_model, substitutes_score=None, threshold=3.0):
def get_substitues(config, tgt_word, keys, atk_model, word_predictions, word_pred_scores_all, top_index, threshold):

    # Get the sub_words and their scores for a particular maksed word
    # substitues L, k
    # from this matrix to recover a word
    substitutes = word_predictions[keys[top_index[0]][0]:keys[top_index[0]][1]]  # L, k
    substitutes_score = word_pred_scores_all[keys[top_index[0]][0]:keys[top_index[0]][1]]

    tokenizer = atk_model['tokenizer']
    mlm_model = atk_model['mlm']

    words = []
    sub_len, k = substitutes.size()  # sub-len, k

    if sub_len == 0:
        return words

    elif sub_len == 1:
        for (i,j) in zip(substitutes[0], substitutes_score[0]):
            if threshold != 0 and j < threshold:
                break
            words.append(tokenizer._convert_id_to_token(int(i)))

        if config['use_ast_constraint'] == 1:
            words = constrain_type(config, tgt_word, words)

    else:
        if config['use_bpe'] == 1:
            words = get_bpe_substitues(config, tgt_word, substitutes, tokenizer, mlm_model)
        else:
            return words
    return words